Support complex-valued functions

[Martin Sandve Alnæs]

Extend UFL to support complex-valued functions with associated operations.

[Prof Garth Wells]

This should be relatively easy from the UFL side, but would require more work on the FFC, UFC and DOLFIN sides to properly support complex types (for feeding into solvers that take complex types).

[Martin Sandve Alnæs]

I'm not so sure it would be easy from the UFL side. Every single symbolic computation done in UFL may have been written under the assumption of real valued numbers, and thus all the code must be reviewed for correctness.

I marked this "priority: trivial" not because it is trivial to implement, but because it is low priority.

[Prof Garth Wells]

Implemented.